Dynamic Mechanisms of Working Memory

 

Haim Sompolinsky

Racah Institute of Physics, Hebrew University and

Center for Brain Research, Harvard University

 

 

Abstract

 

            The ability to maintain information for periods of tens of seconds is essential for many cognitive functions. Electrophysiological measurements suggest that working memory is maintained by persistent electrical activity in cortical neuronal circuits. How such large time scales arise in the neuronal dynamics, and how they are used to faithfully represent memories are among the fundamental open issues in current brain research.  I will present several models that address these issues.

            First, I will discuss the concept of attractor manifolds and its application in network models of spatial working memory. Next, I will consider network models that address the more challenging problem of temporal memories. Lastly, recent experimental evidence for the involvement of single neuron mechanisms of short term memory will be reviewed.

            The dynamic theory of working memory bears on general questions about  the transient response properties of noisy, dynamical systems and their information processing capabilities.